How the software is evolving and where it will go next

Big data image

A new generation of catastrophe risk models are coming online, promising greater transparency and choice. The Oasis Loss Modelling framework, RMS(one) platform, AIR’s Touchstone the Global Earthquake Model and Karen Clark & Company’s (KCC) Characteristic Events are just some of the offerings taking cat modelling to the next level.

For a long time the cat modelling universe was dominated by just three agencies: AIR, RMS and Eqecat. Use of the models took off following Hurricane Andrew in 1992. Initially focused on the “peak zones” of US earthquake and windstorm, European windstorm and Japanese wind and quake, cat modelling has slowly broadened by peril and geography.

As computing power and underlying hazard information has improved, new updated models have been released, sometimes with a significantly altered view of risk. RMS Version 11 US windstorm model for instance, used claims information from events such as Hurricane Ike in 2008 to increase its view of the potential inland damage from hurricanes.

Despite the updates complaints about the models have persisted. The expense, monopoly of the three main vendors and lack of transparency into the assumptions that sit behind the model are causes of frustration.

“Despite considerable refinement of the models over the decades, uncertainty remains – and it is a significantly bigger factor than many users may recognize,” stated Guy Carpenter in one report. “Advances within the modeling industry since 1999 have indeed reduced the width of the uncertainty band, but the consideration of smaller areas of geography only introduces additional uncertainty.”

Each time a major event occurs the models are found wanting in some way. Whether the failure to capture the impact of Hurricane Katrina’s storm surge, Category 2 Ike’s ability to wreak havoc so far inland or the the impact from secondary earthquake perils, questions asked about the level of uncertainty in the models. The providers are keen to point out that models are just tools and not exact approximations of reality.

The lack of models in some parts of the world is another challenge. Major loss events, such as the Thai floods, have occurred in parts of the world that were not modelled. Without tools to manage their aggregations, some insurers have found been disproportionately impacted by these events in so-called cold spots.

Some companies have opted to use their own proprietary models or to blend the output of two or more vendor models. It is believed such an approach can offer the middle ground and buffer the impact of a significant model update, such as RMS v11.

New kids on the block

Now a group of new players is coming onto the scene. The next stage of the journey is about offering greater choice, openness and insight into model uncertainty as well as creating a broad consumer marketplace for risk models.

At Oasis, this year is about integrating a hurricane model into the platform. The Oasis Solutions Project is currently underway at Lloyd’s to test the integration of HurLoss. “This project for Lloyd’s is going to be definitive in how you look at model validation and what transparency practically means in making useful decisions about underwriting and risk selection,” says Oasis project director Dickie Whitaker.

A non-for-profit organisation, Oasis is already punching above its weight with a membership of over 40 re/insurers and brokers. “We are moving very rapidly,” Whitaker continues. “We have accelerated our development which is very challenging and quite exciting but it’s definitely difficult.”

“There are some other new organisations that are growing too like Kat Risk, JBA and Impact Forecasting while the big players like AIR and Core logic EQECAT appear more static,” he adds. “RMS of course have a major shift with the new proposed rollout of RMS ONE.

Oasis has recently partnered with Equinix to provide a hosted solution that Whitaker believes will be more user friendly. “In our world we offer flexibility, which is the ability to customise Oasis,” he explains. “This means you have to do more. So one of the issues for people is which version of the software do they use? There’s lots of choice and getting stuff like that integrated with IT departments is always a challenge, using private hosting providers is one way to avoid that hassle.”

Another open source initiative is the Global Earthquake Model (GEM) Foundation’s OpenQuake model, which officially launched in January. It is the product of a collaboration of over 300 international earthquake experts. The head of the UN Office for Disaster Risk Reduction Margareta Wahlström, hailed the launch as a “major milestone for GEM and the community of earthquake hazard and risk experts”.

Deterministic approaches are another development taking place in catastrophe modelling, particularly for regions and perils that lack probabilistic models. In her product RiskInsight, Clark has taken realistic disaster scenarios to the next level using Characteristic Events (CEs) to look at how a particular event might impact a portfolio.

“The CE approach shows you your losses from the 100-year event, which is almost the flip side of the PML [probable maximum loss],” explains Karen Clark, the founder of KCC. “The value of the CE approach and what it gives companies that the PML approach doesn’t, is a very detailed look at exposure concentrations and hotspots.”

This article first appeared in a GR special report on cloud computing and insurance, in association with Equinix. The full report can be read here.